Re: epics-base/.ci submodule problem

["Zimoch Dirk \(PSI\) via Core-talk" <core-talk at aps.anl.gov>]

When I cloned base 7.0 for my JSONerrors branch a few days ago, I noticed the problem with .ci:
$ git status
...
#	modified:   .ci (new commits, modified content)
and also noticed that the referenced commit did not exist.

$ git submodule update
fatal: reference is not a tree: 8a2666a9deeb6c618bcc73ba80f1a83a878c7a84
Unable to checkout '8a2666a9deeb6c618bcc73ba80f1a83a878c7a84' in submodule path '.ci'

But it did not really hinder my work or stop me from cloning it. It just showed the error.

My git version is 1.8.3.1.

Dirk


On Mon, 2022-08-01 at 10:56 -0700, Michael Davidsaver via Core-talk wrote:
> On 8/1/22 10:51, Andrew Johnson wrote:
> > Thanks, that seems to have fixed it (after blowing away the agent's workspaces again).
> > 
> > No idea why it succeeded on MacOS before that, the git version there is a bit older than the one on the RHEL-7
> > (linux-64) build agent.
> > 
> > Should we be protecting our submodule master branches from force-pushes?
> 
> Yes.  I've been trying to remember to enable protection for
> repositories I've created recently.
> 
> 
> > GitHub can do that now IIRC. A discussion topic for Wednesday I think...
> 
> I've gone ahead and done this for the github.com/epics-base
> repositories I ~regularly push to.  Launchpad remains an open
> items...
> 
> 
> > - Andrew
> > 
> > 
> > On 8/1/22 12:30 PM, Michael Davidsaver wrote:
> > > On 8/1/22 10:21, Andrew Johnson via Core-talk wrote:
> > > > APS Jenkins seems to be unable to clone the epics-base repo (I'm not sure why that just succeeded on macOS,
> > > > which is running exactly the same commands as the Linux agents) due to something in the .ci submodule.
> > > 
> > > It looks like there was a force push to 'master' in ci-scripts
> > > some time on or after the 24th.  ("origin/master" in my checkout
> > > was still pointing to the g8a2666a rev.)
> > > 
> > > So as not to have a dangling submodule reference in the Base history
> > > I have created a tag "v3.3.0-10-g8a2666a" to keep that rev. reachable.
> > > 
> > > 
> > > > On 8/1/22 11:56 AM, APS Jenkins via Core-talk wrote:
> > > > > 
> > > > > Build URL: https://jenkins.aps.anl.gov/job/epics-7.0/OS=linux64/413/
> > > > > 
> > > > > This job failed to build.
> > > > > 
> > > > >  ...
> > > > >  > git config --get submodule.modules/pvDatabase.url # timeout=10
> > > > >  > git config -f .gitmodules --get submodule.modules/pvDatabase.path # timeout=10
> > > > >  > git config --get submodule.modules/pva2pva.url # timeout=10
> > > > >  > git config -f .gitmodules --get submodule.modules/pva2pva.path # timeout=10
> > > > >  > git config --get submodule..ci.url # timeout=10
> > > > >  > git config -f .gitmodules --get submodule..ci.path # timeout=10
> > > > > Setting http proxy: webproxy1.anl.gov:9091
> > > > >  > git submodule update modules/pvData # timeout=10
> > > > > Setting http proxy: webproxy1.anl.gov:9091
> > > > >  > git submodule update modules/pvAccess # timeout=10
> > > > > Setting http proxy: webproxy1.anl.gov:9091
> > > > >  > git submodule update modules/normativeTypes # timeout=10
> > > > > Setting http proxy: webproxy1.anl.gov:9091
> > > > >  > git submodule update modules/pvaClient # timeout=10
> > > > > Setting http proxy: webproxy1.anl.gov:9091
> > > > >  > git submodule update modules/pvDatabase # timeout=10
> > > > > Setting http proxy: webproxy1.anl.gov:9091
> > > > >  > git submodule update modules/pva2pva # timeout=10
> > > > > Setting http proxy: webproxy1.anl.gov:9091
> > > > >  > git submodule update .ci # timeout=10
> > > > > hudson.plugins.git.GitException: Command "git submodule update .ci" returned status code 1:
> > > > > stdout: Cloning into '.ci'...
> > > > > 
> > > > > stderr: fatal: reference is not a tree: 8a2666a9deeb6c618bcc73ba80f1a83a878c7a84
> > > > > Unable to checkout '8a2666a9deeb6c618bcc73ba80f1a83a878c7a84' in submodule path '.ci'
> > > > 
> > > > I'm guessing we may just need to commit a 'git add .ci; git commit' to the 7.0 branch to fix that but I'm not
> > > > certain, can someone explain what happened to the ci submodule?
> > > > 
> > > > - Andrew
> > > > 
> > > > -- 
> > > > Complexity comes for free, Simplicity you have to work for.
> > > > 
> 
>